Research map / Label-efficient and robust learning / Semi-supervised learning
Reducing Spatial Labeling Redundancy for Active Semi-Supervised Crowd Counting
Yongtuo Liu, Sucheng Ren, Liangyu Chai, Hanjie Wu, Dan Xu, Jing Qin, Shengfeng He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| label vertical-strip regions (height equals image height) instead of horizontal strips (ablation) | region-level labeling · ShanghaiTech PartA (10% budget) | MAE: 95.9 -> 89.1 | Improved |
| add MDC region selection instead of random (ablation) | region-level labeling · ShanghaiTech PartA (10% budget) | MAE: 89.1 -> 83.3 | Improved |
| add MDC instead of MAX selection (ablation) | region-level labeling · ShanghaiTech PartA (10% budget) | MAE: 90.8 -> 83.3 | Improved |
| MDC versus random selection when budget is abundant (ablation) | region selection · ShanghaiTech PartA (10% budget) 90% | MAE: 68.7 -> 68.5 | No clear effect |
| add CAP module (train and inference) (ablation) | MDC · ShanghaiTech PartA (10% budget) | MAE: 83.3 -> 78.8 | Improved |
| add CAP module at training only (ablation) | MDC · ShanghaiTech PartA (10% budget) | MAE: 83.3 -> 79.6 | Improved |
| keep CAP at inference versus removing it (ablation) | MDC + CAP · ShanghaiTech PartA (10% budget) | MAE: 78.8 vs 79.6 | No clear effect |
| use 20% warm-up samples instead of 40% (ablation) | MDC · ShanghaiTech PartA (10% budget) | MAE: 78.5 -> 76.4 | Improved |
| use 20% warm-up samples instead of 10% (ablation) | MDC · ShanghaiTech PartA (10% budget) | MAE: 77.1 -> 76.4 | Improved |
| region-level labeling versus None-or-All labeling (ablation) | random selection · ShanghaiTech PartA (10% budget) | MAE | Improved |
| region-level semi-supervised method versus AL-AC | Ours · ShanghaiTech PartA (10% budget) | MAE: 87.9 -> 79.6 | Improved |
| region-level method versus AL-AC on PartB | Ours · ShanghaiTech PartB (10% budget) | MAE: 13.9 -> 12.7 | Improved |
Closest papers by meaning
- Semi-Supervised Crowd Counting via Self-Training on Surrogate Tasks (ECCV 2020 · similarity 0.92)
- Active Crowd Counting with Limited Supervision (ECCV 2020 · similarity 0.91)
- Spatial Uncertainty-Aware Semi-Supervised Crowd Counting (ICCV 2021 · similarity 0.91)
- Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (CVPR 2018 · similarity 0.90)
- Point-to-Region Loss for Semi-Supervised Point-Based Crowd Counting (CVPR 2025 · similarity 0.90)
- Taste More, Taste Better: Diverse Data and Strong Model Boost Semi-Supervised Crowd Counting (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
